GEO audit
A GEO audit you can explain to your client
Five scored pillars, including AI citability. Each 0–100 score comes from concrete checks we publish. Anything we couldn't verify is marked, not guessed.
The audit reads up to 20 pages of the site. A free account (email and password) opens the full report and its audit-based action plan. 3 free audits per connection per day.
Your client asked, "Are we in ChatGPT?"
Nobody can answer that honestly with one number. AI answers change from one day to the next, even when nothing on the site has changed. Here is why, with sources.
What you can answer is the readiness half: can AI assistants fetch this site, read it, and cite it? That is what this audit checks, the same way every time, with every rule written down.
What the audit checks
Five pillars, each scored 0–100. Each one starts at 100. Points come off for findings and for measured quality signals, and every cost is written on the report. The overall score is the weighted sum of the five.
| Pillar | Weight in the overall score | What it looks at |
|---|---|---|
| AI citability | 25% | Whether AI assistants can fetch, parse and cite the site: AI-crawler rules and access, structured data (JSON-LD), readable HTML, question-shaped headings, answer structure, llms.txt. |
| SEO | 25% | Classic search fundamentals, checked on the homepage and across every crawled page. |
| Speed | 15% | Server response time, page weight, compression and mobile setup, measured from the actual fetches. |
| Trust | 20% | TLS, security headers, and the signals buyers look for before they call. |
| Content | 15% | Depth, freshness and presentation of the actual copy, counted from the fetched pages. |
The weights are our judgment, and the methodology says so.
One detail your client may care about: the audit treats two kinds of AI crawler differently. Blocking the crawlers that fetch pages live for answers is a citability problem. Blocking crawlers that only collect training data is a deliberate choice, and the report words it that way.
Every number has a written reason
- Every check, every threshold, every deduction is published. A finding takes off a fixed amount: 35 points for critical, 15 for major, 5 for minor.
- No language model touches a score. Findings and scores come from code. A model only phrases the action plan, and it can't add a finding or move a score.
- Same site, same day, same result. The audit is deterministic, with one stated exception: a single Google PageSpeed Insights lookup in the speed pillar, which can be unavailable. The methodology explains how that is handled.
So when a client asks why the score isn't higher, you can point to the lines on the report that cost the points.
"Couldn't verify" means we didn't guess
Sometimes a check can't collect its evidence. A firewall answers instead of the page, or a probe runs out of budget. When that happens, the report says Couldn't verify. The check doesn't pass and doesn't fail, and it doesn't move the score in either direction.
Here is a real one, from our own site's report:
AI-crawler access: Couldn't verify. No robots.txt rule blocks the AI crawlers, but this audit has no record of fetching the homepage as the AI crawlers. A CDN or firewall block is not ruled out.
That is something you check by hand before you tell a client anything. It is not a problem we invented to fill the report.
What your client sees
Open a sample report. It is the audit of our own site, corpushq.co, run on 2026-10-20.
- Saved reports have a share link. Your client can open it without signing up.
- To share only the AI citability score, add
#citabilityto the report link. - Findings are listed most important first, each with what was checked and a suggested fix.
What this does not do
- It reads up to 20 pages. It is not a complete crawl of every page on a large site.
- It is not a citation tracker. It checks whether AI assistants can read and cite the site. It does not monitor whether ChatGPT or any other assistant actually cites your client.
- It does not fill in or submit forms.
- We do not guarantee rankings or AI mentions.
- A low score means our checks found specific problems, listed with fixes. It does not mean the business is failing.
Checking more than one client
Free: up to 3 saved sites, the full report and audit-based action plan, and a manual re-audit once every 30 days.
Pro: $99 a month per account and covers up to 10 saved sites. It adds a 90-day plan, optional weekly site scans, a chat that answers from the audit, llms.txt, structured-data and crawler-rule drafts, and LinkedIn, X and Reddit-style drafts that you review and publish yourself. Approving a draft never publishes it.
You don't need Pro to see a new audit's full report. Compare Free and Pro.
If a finding is wrong
Send the report link to intake [at] corpuslocalization.com with the subject AUDIT ACCURACY. A person re-runs the check and answers. Because the audit is deterministic, the re-run is evidence, not an opinion.
Questions agencies ask
What is a GEO audit?
A GEO audit checks whether AI assistants such as ChatGPT, Claude and Perplexity can fetch, read and cite a website. Ours scores five pillars from 0 to 100, including AI citability, using checks we publish.
Does it tell me whether ChatGPT cites my client?
No. It checks whether the site is readable and citable, which is the part you can measure the same way twice. Whether an assistant names a business changes from day to day, so we don't sell a single number for it.
How do I explain the score to my client?
Say it is a website audit score, not a count of AI mentions. Each of the five areas starts at 100 and loses points for specific findings, and every deduction is written out on the report. The overall 0–100 score weights SEO and AI citability at 25% each, trust at 20%, and speed and content at 15% each. A check marked couldn’t verify neither helps nor hurts the score.
Can this prove our GEO work caused the improvement?
No. Corpus Plan shows what changed on the website, not why leads, traffic or AI mentions changed. Re-audit after the work and compare: no score moves without a written reason on the report. Free accounts can re-audit once every 30 days; Pro re-audits on request. To connect the work to business results, track where new leads come from separately.
Can my client open the report without an account?
Yes. A saved report has a share link, and anyone with the link can read it without signing up.
Is the GEO audit free?
Yes. The preview needs no account and no card. A free email-and-password account opens the full report and its audit-based action plan.
Is GEO different from SEO?
Mostly, it isn't. AI assistants cite pages they can crawl and understand, so the work starts with ordinary SEO. ChatGPT traffic starts with SEO explains why.
Run it on your next client's site
Start with a free preview, then save it to share the report link.
Read the methodology · Pricing · Free AI visibility checker · ChatGPT traffic starts with SEO